Are you new to machine learning? You heard about Amazon’s machine learning tools, but you’re not sure how to use them? Or perhaps you’re wondering how to get started with AWS machine learning services? This article will help you with these questions by explaining what machine learning is, supplying some use cases of its application, and providing an overview of Amazon’s machine learning tools. Welcome to our compact guide to machine learning at AWS. Let’s answer some fundamental questions first.
1. What is machine learning?
According to Wikipedia, “Machine learning is a field of computer science that gives computers the ability to learn without being explicitly programmed.” It’s an approach to artificial intelligence and also one of the hottest topics in modern IT.
But what does it mean to give computers the ability to learn? How can they learn without being programmed?
Giving computers the ability to learn means feeding them algorithms to learn and adapt based on experience and data. Thanks to these algorithms, computers can make computations and use them to make reliable, repeatable decisions and produce results. They can also adapt when new data is provided.
2. Why do we use machine learning?
Who uses machine learning and for what? There are plenty of use cases for different machine learning algorithms.
Machine learning is instrumental when analyzing vast amounts of historical data to make informed business decisions. By gaining insights from this data, you can get an advantage over your competition, as machine learning algorithms allow you to analyze data in real time to predict future trends.
Almost every industry can benefit from machine learning. For instance, a financial institution can use it to prevent fraud. Health care can assess patients’ health status in real time to identify trends. Marketing and sales can use machine learning to analyze previous purchases and promote items customers may seek to buy.
Machine learning is also very effective at identifying patterns and predicting future trends or malfunctions. The possibilities that machine learning offers are infinite, especially when coupled with cloud services. With relevant input data, model and learning algorithm, one can benefit from machine learning like never before.
For instance, Amazon claims to be using machine learning to predict the demand for bare metal servers they have to order and deploy. Spotify uses machine learning to analyze what music their subscribers listen to so they can come forward with suggested playlists. Machine learning is all over, so why not capitalize on it as Amazon or Spotify do?
3. How can AWS support your machine learning?
As mentioned above, cloud services can support machine learning. AWS offers several services that not only provide support for data analytics, learning models, or computations but can also supply a complete solution, e.g., for voice and image recognition.
Let’s then take a closer look at the tools provided by AWS that you can use for machine learning.
+7 AWS Machine learning tools
1. Amazon Machine Learning
The first machine learning service offered by Amazon is Amazon Machine Learning (yes, no excitement in the name here). It is a service that makes it easy for developers to use machine learning. It provides visualization tools and wizards to guide you through the process of creating machine learning models so that you need no knowledge of machine learning algorithms and technology. When your models are defined, you can quickly obtain predictions using a simple API.
Amazon Machine Learning is highly scalable and can generate billions of real-time predictions every day. With Amazon Machine Learning, you can create a wide variety of predictive applications to forecast demand, personalize content, predict user activity, analyze text and so on. The possibilities are endless.
2. Amazon SageMaker
If you consider the Amazon Machine Learning service too challenging to use, try Amazon SageMaker. It’s one of the newest machine learning services announced by Amazon during the last re:INVENT conference in Las Vegas.
SageMaker is a fully-managed service that enables developers and data scientists to quickly and easily build, train, and deploy machine learning models at any scale.
To use SageMaker, you first need to transform data from other AWS services to Jupyter notebooks running on fully-managed instances, and include code for the common model training and hosting exercises. When that transformation is complete, you can use SageMaker algorithms and frameworks or bring your ones to train your model. Just indicate the type and quantity of instances needed and start the training with one click. After the training, SageMaker will adjust multiple combinations of algorithm parameters to tune up your model. A completed training can be deployed for real-time predictions.
A good thing about SageMaker is that you don’t need any in-depth knowledge of machine learning to prepare, train, and tune machine learning model and start predictions. Typical use cases for SageMaker are optimizing your return on ad spend, predicting the likelihood of credit default, or real-time prediction to anticipate machinery failure or maintenance schedule.
Apart from Amazon Machine Learning and SageMaker, Amazon provides several tools available as a service. All of them can be used as a part of your application.
3. Amazon Comprehend
Amazon Comprehend is a continuously-trained Natural Language Processing (NLP) service. With Comprehend, you can achieve two significant things:
- The first is to perform NLP tasks like detecting the dominant language in a document, detecting entities such as persons or places in a document, detecting sentiment and key phrases.
- The second is to examine a collection of documents to determine common themes such as sports, politics, or entertainment.
4. AWS DeepLens
AWS DeepLens is another service announced at the last re:INVENT. It is a deep learning-enabled video camera that you can use for face, object, and action recognition or for applying an artistic style to images. The service runs deep learning directly on a device, so data doesn’t have to move to the cloud for processing. It also integrates with other AWS services, such as Amazon SageMaker, to allow you to build and train models in the cloud quickly.
5. Amazon LEX
LEX is a bot you can deploy and train in the cloud to interact with your customers. LEX bot can integrate with Amazon Connect to handle a chat. The service can recognize the voice and transfer it to text for further processing. LEX can be attached to third-party services such as Slack or Facebook, or it may provide services such as reservations, for instance.
6. Amazon Polly
Polly is a service that you can integrate with your applications and services to provide Text-to-Speech conversion. For example, you can use it with Lambda to save text to other formats, e.g., mp3. Polly supports 47 voices in 24 languages.
7. Amazon Rekognition
Finally, there’s Amazon Rekognition. It is a deep learning-based visual analysis service, which enables you to search, verify, and organize millions of images and videos. Rekognition is continuously learning from aggregating data. It can detect object and scenes, moderate images, perform facial analysis, face comparison, celebrity recognition, recognize text in images and analyze videos.
Amazon provides SDK and API to ease the usage of Rekognition. Besides that, Rekognition seamlessly integrates with Amazon S3 and Lambda, so that your applications can analyze data directly from S3. Real-time analysis is also available for Amazon Kinesis Video Streams.
More: Amazon Transcribe & Translate
Two services are currently in a preview phase, these are Amazon Transcribe, for automatic speech recognition, and Amazon Translate, which is a neural machine translation service. They are to go public soon.
Machine learning – endless possibilities
Machine learning is a technology of the future, and soon it will be present in every aspect of our lives. Even if you have no in-depth knowledge of it, you will be able to take advantage of machine learning with the tools provided by the cloud. With Amazon services, you should be able to create the desired machine learning model and applications. Use them with caution, as with great power comes great responsibility.